Heat Vulnerability Assessment in the Boston Metropolitan Area

Suitability Analysis in Python

Rachel Bowers | May 11, 2021

Final project for UEP239: Geospatial Programming with Python

Project Overview

Extreme heat can be a public health risk, especially to vulnerable populations like young children, elderly, and low-income individuals. Climate change and urbanization are increasing the frequency and, in some areas, the intensity of extreme heat events. This analysis uses Python to provide a spatial understanding for which Boston Metropolitan area communities (defined as Zip Code Tabulation Areas (ZCTAs) within the Boston Region Metropolitan Planning Organization (MPO) boundary) are most vulnerable to extreme heat events.

Amongst other characteristics, areas with high land surface temperature, low vegetation, highly vulnerable populations, and limited access to urgent care hospitals are particularly at risk of the health issues associated with extreme heat exposure (Reid Colleen E. et al. 2009). To capture these characteristics, this analysis defines extreme heat vulnerability based on the following indicators:

Analysis

First, import dependencies

Study Area

Clip ZCTA polygons to the MPO boundary

There are 164 ZCTAs in the Boston MPO region.

Preparing Indicators

Indicator: Age (Under 5 & Over 65)

Young and old individuals are more vulnerable to the health impacts associated with extreme heat exposure (Reid Colleen E. et al. 2009). To capture this, I used census data to calculate percentages of populations under the age of 5 and over the age of 65 in Boston area ZCTAs.

Join age dataframe to ZCTA geodataframe

Age Under 5: Most vulnerable ZCTAs

ZCTAs with higher populations under 5 appear to be randomly dispersed throughout the Boston MPO. ZCTAs in Fayville, Hanscom, Charlestown, and Chelsea score highest for this indicator.

Age Over 65: Most vulnerable ZCTAs

ZCTAs with higher populations over the age of 65 are concentrated in the inner ring suburbs and North Shore areas, like Humarock, Rockport, Boston, Lincoln, and Nahant.

Indicator: Income

Low-income populations tend to have less resources for keeping their homes cool in warm months, making them more vulnerable to extreme heat (Reid Colleen E. et al. 2009). To capture this, I used median household income estimates from the American Community Survey. ZCTAs with lower median household incomes will be considered more vulnerable in this analysis.

Income: Most vulnerable ZCTAs

Income: Least vulnerable ZCTAs

Vulnerable ZCTAs are ZCTAs where median household income is lowest. This is true in ZCTAs located in Lynn, Roxbury, Dorchester, Roxbury Crossing, and Boston. Least vulnerable (highest income) ZCTAs are in Waban, Wellesley Hills, Weston, Sherborn, and Carlisle.

Indicator: Land Surface Temperature

Land surface temperature, derived from Landsat 8 satellite imagery, is a proxy for estimating urban heat island effect (“Climate Vulnerability in Greater Boston”). Higher land surface temperature indicates higher vulnerability to extreme heat. The imagery used in this analysis was taken on a cloudless day on May 21, 2020.

Scale, Offset, Convert to Celsius

From Documentation: "A scale factor must be applied to both Collection 1 and Collection 2 Landsat Level-2 surface reflectance and surface temperature products before using the data. Landsat Collection 2 surface temperature has a scale factor of 0.00341802 and an additional offset of 149.0 per pixel."

Land Surface Temperature: Most vulnerable ZCTAs

Land Surface Temperature: Least vulnerable ZCTAs

ZCTAs with higher average land surface temperature are clustered toward Boston's urban core in Boston, Somerville, Allston, Chelsea, and Cambridge, with some outlying areas having higher average land surface temperature than their surroundings. ZCTAs with lowest average land surface temperature are found in Manchester, Rockport, Essex, Norwell, and Dover.

Indicator: Vegetation (Normalized Difference Vegetation Index (NDVI))

The Normalized Differential Vegetation Index (NDVI) is a standardized vegetation index which is calculated using near infra-red (band 5) and red (band 4) bands. Vegetation provides shade and cooling effects. Areas with lower vegetation (or lower NDVI) are more vulnerable to health impacts from extreme heat (Wasser 2019).

Vegetation: Most vulnerable ZCTAs

Vegetation: Least vulnerable ZCTAs

Vegetation follows a similar pattern as Land Surface Temperature, with areas closer to the urban core experiencing less vegetation than outlying suburbs. ZCTAs located in Boston, Lynn, and Cambridge have the lowest NDVI values. ZCTAs in Bolton, Sherborn, Dover, Carlisle, and Weston have the highest NDVI values.

Indicator: Hospital Density

Easy access to hospitals is important for getting treatment for health issues that arise from exposure to extreme heat (Reid Colleen E. et al. 2009). To represent hospital access, this analysis uses count of acute care hospitals (from MassGIS) per ZCTA.

Hospital Count: Least vulnerable ZCTAs

ZCTAs witout urgent care hospitals are the most vulnerable, yet are also the most common. The least vulnerable ZCTAs (areas with the most urgent care hospitals) can be found in Boston.

Vulnerability Model

Now that all of the indicator variables are prepared, let's put it all together in a vulnerability model.

Weighted Vulnerability: I chose to set land surface temperature as 50% of heat vulnerability, and all other indicators equally at 10% each.

Interactive Map and Results Discussion

Weighted Heat Vulnerability Analysis: Most vulnerable ZCTAs

Weighted Heat Vulnerability Analysis: Least vulnerable ZCTAs

Findings

Based on this analysis, the ZCTAs where communities are most vulnerable to extreme heat issues are located in Boston, Chelsea, Somerville, Lynn, Cambridge, Roxbury, and Allston. The ZCTAs where communities are least vulnerable to extreme heat issues are located in Manchester, Sherborn, Carlisle, Rockport, and Norwell.

Next Steps

As climate change results in intensified extreme heat events, identifying policy and planning interventions to address heat vulnerability will be important, particularly in communities located in this analysis. This study would be enhanced with more granular data to identify blocks that are particularly vulnerable to extreme heat, as well as with additional data about things like building height (a proxy for indoor heat concentration), air conditioning, and health data.

References